Generalized Processor Sharing(GPS) has been considered as an ideal scheduling discipline based on its fairness properties and worst-case delay bound. Until recently, emulating of GPS has been regarded as the ideal means of providing fairness properties and worst-case delay bound. These are WFQ(Weighted Fair Queueing), SCFQ(Self-Clocked Fair Queueing) and WF2Q(Worst-case Fair Weighted Fair Queueing).
But they have not considered delay and mean delay seriously, and hence even the queue delay of GPS discipline is larger than that of FIFO. The larger delay makes the queue be overflowed rapidly, and subsequently many cells(or packets) are lost by queue. Cell Loss Ratio(CLR) affects the quality of realtime service, and therefor it is not desirable to maintain cell loss ratio high.
In this paper, we present a method for the design of scheduling algorithm that provides the low queue delay and also low mean queue delay. Instead of emulating GPS, we propose a new scheduling scheme which deals with traffic by a rate proportional delay. Since this scheme services the maximum delayed packet in the HOL(Head Of Line) and longest session first, it can improve the queue delay with respect to that of GPS. Also, since proposed algorithm makes each session dependent, it can cooperate with buffer management algorithm as RED(Random Early Detection) method. Because random packet drop by RED in one queue makes serious impact on the delay of all queues. We compare the delay performance of the proposed scheme with that of fair queueing by mathematical analysis and computer simulation results.
